Season 1 is complete. The last round ended 24 September 2026 at 09:00 UTC. This is the permanent archive: every round, every trader, every inherited change.
Gold thread: a trader carried into the next round, thicker the longer it has survived. Each elder sits just inside its ring with its eight mutants fanned outside. Wash: family territory. Drag to pan, scroll to zoom, click a dot to trace its bloodline.
Round 41final

Final round · Round 41 · 08:00–09:00 UTC

Who made money last round?

All 45 traders of the last completed round. Before fees decides who survives; after fees is what a real account would have kept.

Rank by
#TraderChange from parentBefore feesTradesAfter feesResult
1BEcho · Bold · 5-min View · Sees 1m/10m/1h/6h +2v3-g0041-p2-m7 · up to 3.0xDecides: every 30 min → every 3 min+$32.573 (1 at the bell)−$11.87survived
2BEcho · Bold · 5-min View · Sees 10m/1h/3h/24h +1v3-g0041-p2-m4 · up to 3.0xWhat it sees: 1m · 10m · 1h · 6h → 10m · 1h · 3h · 24hHow it reads facts: numbers only → in words+$4.371 (1 at the bell)−$10.35survived
3BEcho · Bold · 10-min View · Sees 1m/10m/1h/6h +2v3-g0041-p2-m5 · up to 3.0xForecast horizon: 5 min → 10 min+$4.371 (1 at the bell)−$10.35survived
4BEcho · Eager · 5-min View · Sees 1m/10m/1h/6h +2v3-g0041-p2-m6 · up to 3.0xRisk stance: Bold → Eager Prioritize identifying and acting on the supplied trading opportunities.+$4.371 (1 at the bell)−$10.35survived
5BEcho · Bold · 5-min View · Sees 10m/1h/3h/24h +2v3-g0041-p2-m3 · up to 3.0xWhat it sees: 1m · 10m · 1h · 6h → 10m · 1h · 3h · 24h+$4.371 (1 at the bell)−$10.35survived
6BEcho · Homing Pigeon · 30-min View · Sees 15m/90m/6h/24h +1v3-g0040-p1-m4 · parent, unchanged · up to 0.3xunchanged+$2.471 (1 at the bell)+$1.23eliminated
7BEcho · Homing Pigeon · Sees 10m/1h/3h/24h · Every 30 minv3-g0039-p3-m3 · parent, unchanged · up to 0.3xunchanged+$1.902 (1 at the bell)−$0.59eliminated
8BEcho · Homing Pigeon · 30-min View · Sees 15m/90m/6h/24h +2v3-g0040-p1-m3 · parent, unchanged · up to 0.3xunchanged+$1.902 (1 at the bell)−$0.59eliminated
9BEcho · Trend Rider · Lean · 5-min View +2v3-g0041-p1-m8 · up to 1.0xHypothesis: Homing Pigeon → Trend Rider Evaluate whether the recent price movement is likely to continue.Forecast horizon: 30 min → 5 min+$1.471 (1 at the bell)−$3.49eliminated
10BEcho · Homing Pigeon · Lean · 30-min View +2v3-g0041-p1-m3 · up to 0.3xWhat it sees: 10m · 30m · 3h · 12h → 15m · 90m · 6h · 24h+$1.252 (1 at the bell)−$1.24eliminated
11BEcho · Homing Pigeon · Lean · 30-min View +3v3-g0041-p1-m4 · up to 0.3xWhat it sees: 10m · 30m · 3h · 12h → 1m · 10m · 1h · 6hHow it reads facts: in words → numbers only+$1.252 (1 at the bell)−$1.24eliminated
12BEcho · Homing Pigeon · Bold · 5-min View +3v3-g0041-p2-m2 · up to 0.3xHypothesis: Open Book → Homing Pigeon Evaluate whether price is likely to return toward its observed range centre.+$1.252 (1 at the bell)−$1.24eliminated
13BEcho · Homing Pigeon · Sees 5m/15m/1h/4h · Numbers Only +1v3-g0041-p3-m4 · up to 0.3xWhat it sees: 10m · 1h · 3h · 24h → 5m · 15m · 1h · 4hHow it reads facts: in words → numbers only+$1.082 (1 at the bell)−$1.41eliminated
14BEcho · Homing Pigeon · 30-min View · Sees 15m/1h/6h/24h +2v3-g0041-p4-m3 · up to 0.3xWhat it sees: 15m · 90m · 6h · 24h → 15m · 1h · 6h · 24h+$1.062 (1 at the bell)−$1.43eliminated
15BEcho · Homing Pigeon · 30-min View · Sees 15m/1h/6h/24h +1v3-g0041-p4-m4 · up to 0.3xWhat it sees: 15m · 90m · 6h · 24h → 15m · 1h · 6h · 24hHow it reads facts: numbers only → in words+$0.722 (1 at the bell)−$1.78eliminated
16BEcho · Top Caller · 5-min View · Sees 10m/1h/3h/24h +1v3-g0041-p3-m8 · up to 0.3xHypothesis: Homing Pigeon → Top Caller Evaluate whether the recent price movement is exhausting and may reverse.Forecast horizon: 15 min → 5 min+$0.371 (1 at the bell)−$0.88eliminated
17BEcho · Trend Rider · 30-min View · Sees 15m/90m/6h/24h +2v3-g0041-p4-m1 · up to 0.3xHypothesis: Homing Pigeon → Trend Rider Evaluate whether the recent price movement is likely to continue.+$0.371 (1 at the bell)−$0.88eliminated
18BEcho · Bold · 5-min View · Sees 1m/10m/1h/6h +2v3-g0040-p4-m1 · parent, unchanged · up to 3.0xunchanged+$0.161 (1 at the bell)−$14.57eliminated
19BEcho · Wanderer · Lean · 30-min View +2v3-g0041-p1-m1 · up to 1.0xHypothesis: Homing Pigeon → Wanderer Seek trading opportunities across the different supplied market contexts.+$0.061 (1 at the bell)−$4.91eliminated
20BEcho · Tie Breaker · 30-min View · Sees 15m/90m/6h/24h +1v3-g0041-p5-m2 · up to 1.0xHypothesis: Homing Pigeon → Tie Breaker Evaluate disagreement between recent and broader price movements.+$0.061 (1 at the bell)−$4.91eliminated
21BEcho · Top Caller · 20-min View · Sees 15m/90m/6h/24h +2v3-g0041-p4-m8 · up to 0.3xHypothesis: Homing Pigeon → Top Caller Evaluate whether the recent price movement is exhausting and may reverse.Forecast horizon: 30 min → 20 min+$0.011 (1 at the bell)−$1.23eliminated
22BEcho · Breakout Chaser · 30-min View · Sees 15m/90m/6h/24h +1v3-g0041-p5-m1 · up to 0.3xHypothesis: Homing Pigeon → Breakout Chaser Evaluate whether movement beyond the recent range is likely to persist.+$0.011 (1 at the bell)−$1.23eliminated
23BEcho · Homing Pigeon · Careful · Sees 10m/1h/3h/24h +1v3-g0041-p3-m6Risk stance: Even Keel → Careful Give particular attention to how adverse movement could affect capital.$0.000$0.00eliminated
24BEcho · Homing Pigeon · Lean · 30-min View +2v3-g0040-p1-m6 · parent, unchanged · up to 0.3xunchanged−$0.361 (1 at the bell)−$1.61eliminated
25BEcho · Homing Pigeon · Bold · 30-min View +2v3-g0041-p5-m6 · up to 0.3xRisk stance: Even Keel → Bold Give particular attention to the attainable upside of the supplied opportunity.−$0.361 (1 at the bell)−$1.61eliminated
26BEcho · Homing Pigeon · 5-min View · Sees 15m/90m/6h/24h +1v3-g0041-p5-m8 · up to 0.3xForecast horizon: 30 min → 5 min−$0.361 (1 at the bell)−$1.61eliminated
27BEcho · Homing Pigeon · 5-min View · Sees 15m/90m/6h/24h +2v3-g0041-p4-m5 · up to 0.3xForecast horizon: 30 min → 5 min−$0.361 (1 at the bell)−$1.61eliminated
28BEcho · Homing Pigeon · Lean · 30-min View +3v3-g0041-p4-m6 · up to 0.3xRisk stance: Even Keel → Lean Evaluate the return opportunity relative to the amount of capital exposed.−$0.361 (1 at the bell)−$1.61eliminated
29BEcho · Homing Pigeon · 30-min View · Sees 1m/5m/30m/2h +1v3-g0041-p5-m3 · up to 0.3xWhat it sees: 15m · 90m · 6h · 24h → 1m · 5m · 30m · 2h−$0.361 (1 at the bell)−$1.61eliminated
30BEcho · Homing Pigeon · 30-min View · Sees 5m/1h/4h/12h +2v3-g0041-p5-m4 · up to 0.3xWhat it sees: 15m · 90m · 6h · 24h → 5m · 1h · 4h · 12hHow it reads facts: in words → numbers only−$0.361 (1 at the bell)−$1.61eliminated
31BEcho · Homing Pigeon · 20-min View · Sees 15m/90m/6h/24h +1v3-g0041-p5-m5 · up to 0.3xForecast horizon: 30 min → 20 min−$0.361 (1 at the bell)−$1.61eliminated
32BEcho · Homing Pigeon · Lean · 20-min View +2v3-g0041-p1-m5 · up to 0.3xForecast horizon: 30 min → 20 min−$0.691 (1 at the bell)−$1.94eliminated
33BEcho · Homing Pigeon · Careful · 30-min View +2v3-g0041-p1-m6 · up to 0.3xRisk stance: Lean → Careful Give particular attention to how adverse movement could affect capital.−$0.691 (1 at the bell)−$1.94eliminated
34BEcho · Homing Pigeon · 20-min View · Sees 10m/1h/3h/24h +1v3-g0041-p3-m5 · up to 0.3xForecast horizon: 15 min → 20 min−$0.691 (1 at the bell)−$1.94eliminated
35BEcho · Top Caller · Bold · 5-min View +3v3-g0041-p2-m1 · up to 0.3xHypothesis: Open Book → Top Caller Evaluate whether the recent price movement is exhausting and may reverse.−$1.892 (1 at the bell)−$4.38eliminated
36BEcho · Homing Pigeon · Lean · 30-min View +2v3-g0041-p1-m7 · up to 0.3xDecides: every 30 min → every 1 min−$2.223 (1 at the bell)−$5.97eliminated
37BEcho · Homing Pigeon · 30-min View · Sees 15m/90m/6h/24hv3-g0041-p5-m7 · up to 0.3xDecides: every 30 min → every 5 min−$2.562 (1 at the bell)−$5.07eliminated
38BEcho · Homing Pigeon · 30-min View · Sees 15m/90m/6h/24h +2v3-g0041-p4-m7 · up to 0.3xDecides: every 30 min → every 1 min−$2.572 (1 at the bell)−$5.07eliminated
39BEcho · Homing Pigeon · Sees 10m/1h/3h/24h · Every 15 minv3-g0041-p3-m7 · up to 0.3xDecides: every 30 min → every 15 min−$2.651 (1 at the bell)−$3.90eliminated
40BEcho · Top Caller · 30-min View · Sees 15m/90m/6h/24h +2v3-g0041-p4-m2 · up to 0.5xHypothesis: Homing Pigeon → Top Caller Evaluate whether the recent price movement is exhausting and may reverse.−$3.492 (1 at the bell)−$8.45eliminated
41BEcho · Homing Pigeon · Every 30 minv3-g0041-p3-m3 · up to 0.5xWhat it sees: 10m · 1h · 3h · 24h → 5m · 30m · 2h · 24h−$5.301 (1 at the bell)−$7.79eliminated
42BEcho · Trend Rider · Bold · 20-min View +3v3-g0041-p2-m8 · up to 1.0xHypothesis: Open Book → Trend Rider Evaluate whether the recent price movement is likely to continue.Forecast horizon: 5 min → 20 min−$7.022 (1 at the bell)−$16.87eliminated
43BEcho · Tie Breaker · Sees 10m/1h/3h/24h · Every 30 minv3-g0041-p3-m1 · up to 1.0xHypothesis: Homing Pigeon → Tie Breaker Evaluate disagreement between recent and broader price movements.−$7.022 (1 at the bell)−$16.87eliminated
44BEcho · Wanderer · Sees 10m/1h/3h/24h · Every 30 minv3-g0041-p3-m2 · up to 1.0xHypothesis: Homing Pigeon → Wanderer Seek trading opportunities across the different supplied market contexts.−$7.022 (1 at the bell)−$16.87eliminated
45BEcho · Lean · 30-min View · Sees 10m/30m/3h/12h +1v3-g0041-p1-m2 · up to 1.0xHypothesis: Homing Pigeon → Open Book Find trading opportunities in the supplied price movements.−$7.552 (1 at the bell)−$17.40eliminated

Full round page →

Who is behind this

Solberg Invest is an open research desk.

Independent research on crypto and equities from Norway, written by named people who show their work: Oskar on macro, monetary policy, crypto and fundamentals, Louis on technical analysis, Simen on Nordic equities, portfolio structure and risk. Read by more than 11,000 investors. The Lab is what the desk does when it would rather test an idea in public than argue about it.

We NEVER run ads.

We NEVER post affiliate links.

We NEVER get paid to promote coins or stocks.

Read about Solberg Invest
Oskar Solberg speaking at an AI conference
Oskar Solberg, founder. The Lab is his experiment; the desk publishes it whether it works or not.

Last round's duel

Did the mutation help?

Last round each parent ran unchanged beside its eight mutants. Same market, same hour. Each mutant changed one inherited trait; compare before and after below.

BEcho · Homing Pigeon · Lean · 30-min View +2parent

v3-g0040-p1-m6 · ran unchanged in round 41 · eliminated this round
Homing Pigeon · Lean · 30-min view · sees 10m · 30m · 3h · 12h · in words · every 30 min · no memory
Before fees
−$0.36
1 trade
After fees
−$1.61
fees $1.25
Exposure
up to 0.3x
of $1,000
−$8−$6−$4−$2$0+$2Parent −$0.36Mutant m2: −$7.55 before fees on 2 trades2Mutant m7: −$2.22 before fees on 3 trades7Mutant m5: −$0.69 before fees on 1 trade5Mutant m6: −$0.69 before fees on 1 trade6Mutant m1: +$0.06 before fees on 1 trade1Mutant m3: +$1.25 before fees on 2 trades3Mutant m4: +$1.25 before fees on 2 trades4Mutant m8: +$1.47 before fees on 1 trade8better than parent →
−$8−$6−$4−$2$0+$2Parent −$0.36Mutant m2: −$7.55 before fees on 2 trades2Mutant m7: −$2.22 before fees on 3 trades7Mutant m5: −$0.69 before fees on 1 trade5Mutant m6: −$0.69 before fees on 1 trade6Mutant m1: +$0.06 before fees on 1 trade1Mutant m3: +$1.25 before fees on 2 trades3Mutant m4: +$1.25 before fees on 2 trades4Mutant m8: +$1.47 before fees on 1 trade8better than parent →

Before-fee result of each mutant (numbered dots) against the parent (gold line), 08:00–09:00 UTC. Filled dots beat the parent, hollow did not, dashed completed no trades.

4 of 8 mutants did better before fees than their unchanged parent in the same hour. Neither the parent nor any mutant survived.One hour is one observation. The useful question is whether the same change keeps helping in later rounds.
  1. m1
    WandererHypothesis: Homing Pigeon → Wanderer Seek trading opportunities across the different supplied market contexts.
    1 trade · 1 closed at the bell · up to 1.0xafter fees −$4.91
    +$0.06▲ +$0.42 vs parent
  2. m2
    Open BookHypothesis: Homing Pigeon → Open Book Find trading opportunities in the supplied price movements.
    2 trades · 1 closed at the bell · up to 1.0xafter fees −$17.40
    −$7.55▼ −$7.19 vs parent
  3. m3
    Sees 15m/90m/6h/24hWhat it sees: 10m · 30m · 3h · 12h → 15m · 90m · 6h · 24h
    2 trades · 1 closed at the bell · up to 0.3xafter fees −$1.24
    +$1.25▲ +$1.61 vs parent
  4. m4
    Numbers OnlyWhat it sees: 10m · 30m · 3h · 12h → 1m · 10m · 1h · 6hHow it reads facts: in words → numbers only
    2 trades · 1 closed at the bell · up to 0.3xafter fees −$1.24
    +$1.25▲ +$1.61 vs parent
  5. m5
    20-min ViewForecast horizon: 30 min → 20 min
    1 trade · 1 closed at the bell · up to 0.3xafter fees −$1.94
    −$0.69▼ −$0.33 vs parent
  6. m6
    CarefulRisk stance: Lean → Careful Give particular attention to how adverse movement could affect capital.
    1 trade · 1 closed at the bell · up to 0.3xafter fees −$1.94
    −$0.69▼ −$0.33 vs parent
  7. m7
    Every 1 minDecides: every 30 min → every 1 min
    3 trades · 1 closed at the bell · up to 0.3xafter fees −$5.97
    −$2.22▼ −$1.85 vs parent
  8. m8
    5-min ViewHypothesis: Homing Pigeon → Trend Rider Evaluate whether the recent price movement is likely to continue.Forecast horizon: 30 min → 5 min
    1 trade · 1 closed at the bell · up to 1.0xafter fees −$3.49
    +$1.47▲ +$1.84 vs parent
The parent's full instruction
Rendered from the parent's inherited traits. A mutant changes one of them; its before and after are shown above.Grow your simulated balance by finding and taking promising BTC trades. Choose the BTC position direction for the best before-fee portfolio return over the next 30 minutes, using the supplied facts, portfolio and experiences. Evaluate whether price is likely to return toward its observed range centre. Evaluate the return opportunity relative to the amount of capital exposed. Spread and slippage affect returns; fees remain payable but are excluded from the evaluation objective. The forecast horizon is not a required holding time.

The game

One model. 45 instruction sets. One hour.

1 · Trade

An hour on the market

Each trader starts the hour with $1,000 of simulated cash. Every one to thirty minutes it answers long, short, flat or keep, and picks a total exposure from 25% to 300% of its equity: up to 3x, in either direction.

2 · Score

The bell rings

At the top of the hour every position is closed. The score is the profit or loss before fees, including that forced close. Fees of 0.25% per fill and 0.02% slippage are charged anyway and shown as after fees.

3 · Survive

Five live, forty die

The five best before-fee results continue unchanged. Everyone else is eliminated and kept in the archive. A 2% move against an entry forces an exit; a portfolio that falls to 20% of its notional is liquidated.

4 · Breed

Eight mutants each

Each mutant inherits its parent's instruction and changes one trait: the hypothesis it tests, its risk stance, its forecast horizon, the windows it sees, how it reads the facts, or how often it decides. Its name lists what differs from its founder. No weights train and nothing is remembered between rounds.

The question: can selecting on inherited traits alone produce traders that keep doing better than the ones they came from? Survival shows who was best in one hour. The evidence is what happens in the next one, and we publish it either way.

Field notes

What Oskar is seeing

Dated notes that separate what the record shows, what Jev said, and what Oskar makes of it.

Oskar Solberg · after round 41Round 41

Parents are not safe, and the Echoes are spreading

Observed
Round 41 ran 08:00–09:00 UTC. 10 of 40 mutants did better before fees than their own unchanged parent. 5 of the five survivors were born this round. The best result before fees, +$32.57 from Echo · Bold · 5-min View · Sees 1m/10m/1h/6h +2 (v3-g0041-p2-m7), came from 3 completed trades at up to 3.0x; after fees it was −$11.87. The Echoes now make up 45 of the 45 traders.
Jev said
Model explanations are quoted here once Oskar has read the ledger of the round. They are the own text of the model, not evidence that a sentence caused a result.
My read
One hour is one draw. The test is whether the same inherited change keeps its mutants above their parents in the next rounds. If it does not, this was noise, and the page will say so.

Caveats

Before you read too much into it

Is this real money?+
No. Every trader has a simulated $1,000 portfolio. Prices, spread, slippage and fees are simulated from a real market feed, but no order is ever sent to an exchange.
What do the two numbers mean?+
Before fees decides survival: the profit or loss of the hour before fees, including the forced close when the bell rings. After fees is the real simulated ledger, with 0.25% per fill and slippage. A 3x round trip on $1,000 costs about $15 in fees, so a survivor with a small positive score has usually lost money after fees. We show both because the score alone would flatter that.
Why can a survivor have lost money?+
Survival means top five of 45 for one hour, before fees. In a bad hour the fifth-best result is negative, and a positive result before fees can turn negative after them. Traders that completed no trades are shown as such rather than counted as evidence either way.
Is the leverage real?+
It is synthetic. The paper broker behind the price feed offers neither short selling nor margin for crypto, so the lab simulates signed exposure up to 3x itself, with zero borrow and funding cost, a 2% safety exit and liquidation at 20% maintenance margin. Treat leveraged results as an assumption about execution, not a replica of an exchange.
Does the AI actually learn?+
Not in the usual sense. Model weights never change and this season the traders keep no memory between rounds. What changes is the instruction, one inherited trait per mutant: the hypothesis it tests, its risk stance, its forecast horizon, what it sees, how it reads the facts or how often it decides. The question is whether selection on those traits alone produces an improvement that holds up.
How fresh is this page?+
The experiment publishes a read-only snapshot about once a minute. If a snapshot is late, the page says so and shows the last verified one with its time. It never shows stale numbers as live.